Comment by ProjectArcturis
2 days ago
This is probably an uninformed question, but why are you comparing layers' output to ground truth? Isn't the point that ground truth is unknown?
2 days ago
This is probably an uninformed question, but why are you comparing layers' output to ground truth? Isn't the point that ground truth is unknown?
During neural network training, the ground truth (GT) must be known to compute the loss.
In DDN, the GT is only used to calculate the loss and guide sampling; it never becomes an input to the model.